基于混合神经进化算法的风电爬坡分类预测

Wind Power Ramp Events Classification Prediction Based on Hybrid Neural Evolution Algorithm

针对风电场风电功率爬坡事件预测问题,提出一种分类预测模型。该模型将大气再分析数据输入的预测问题建模为二分类问题。另一方面,提出一种混合神经进化算法,将极限学习机等人工神经网络与进化算法相结合,对训练后的模型进行优化。由于风电爬坡事件发生的概率很低,因此是不平衡的。本文采用SMOTE采样方法,通过减少大多数类的数据样本来获得良好的结果。采用进化算法进行特征选择,以选择最优的训练模型。最后,通过算例仿真对模型进行评价,验证了模型的精度性能。结果表明,该方法在分类预测性能上具有较好的精度。

Aiming at the problem of wind power climbing event prediction in wind farm,a classification prediction model is proposed.The prediction problem of atmospheric reanalysis data input is modeled as a binary classification problem.On the other hand,this paper proposes a hybrid neural evolutionary algorithm,the limit of machine learning and artificial neural network combined with evolution algorithm,optimize the model after the training.Because the wind climbing event probability is very low,so it is not balanced.In this paper,SMOTE sampling method is adopted to obtain good results by reducing the samples of most classes.Evolutionary algorithm is used to select the best training model.Finally,the model is evaluated by numerical simulation,and the accuracy of the model is verified.The results show that the method has better accuracy in classification prediction performance.